Convective parameterization of ensemble weighted approach for the
Transcrição
Convective parameterization of ensemble weighted approach for the
CONVECTIVE PARAMETERIZATION OF ENSEMBLE WEIGHTED APPROACH FOR THE REGIONAL MODEL BRAMS Ariane Frassoni1, Eduardo F. P. Luz3, Saulo R. Freitas1, Haroldo F. de Campos Velho2, Manoel A. Gan1, João G. Z. De Mattos1, Georg Grell4 [email protected] 1-Center for Weather Forecasts and Climate Studies, National Institute for Space Research, C. Paulista, SP, Brazil 2-Laboratory of Computing and Applied Mathematics, National Institute for Space Research, S. J. Campos, SP, Brazil 3-National Center for Monitoring and Early Warning of Natural Disasters 4-Department of Commerce, National Oceanic and Atmospheric Administration, Earth System Research Laboratory, Global System Division, Boulder, CO, Estados Unidos RESUMO: No presente trabalho, a metodologia de problema inverso de estimação de parâmetros é aplicada ao modelo BRAMS. O problema inverso é resolvido pelo método de otimização Firefly (FA) e o modelo direto é dado pelas simulações de precipitação utilizando diferentes parametrizações de convecção do modelo. O objetivo é determinar numericamente os pesos de cada parametrização para ponderar o conjunto de parametrizações convectivas. Como resultado, é obtido o campo de chuva reconstruído a partir da combinação entre as simulações e os campos de pesos. Os resultados indicaram um campo de precipitação reconstruído mais próxima das observações. ABSTRACT: In this study, the methodology of inverse problem of parameter estimation is applied using the regional model BRAMS. The inverse problem is solved by the Firefly (FA) optimization method and the direct model is given by the simulations of precipitation expressed by different cumulus parameterizations of the model. The goal is to determine numerically the weights of each parameterization to weight the ensemble of cumulus parameterizations. The results showed a retrieved precipitation field closest to the observations. METHODOLOGY RESULTS INTRODUCTION Why is so difficult represent clouds in the numerical models? Experimental tests with the FA algorithm parameter Parameter Initial Value Final Value Increment α 0.01 0.1 0.01 β 0.1 1.0 0.1 γ 1.0 10. 1.0 n 5 5 5 G 10 100 10 Weights estimation Inverse Problems Direct problem Observation s (effects) Parameters (causes) Inverse problem GCM Grid cell 40400km The model BRAMS Categorical verification: Performance Diagram to threshold 50.8 mm for South America, for January 2008. AS, EN, GR, LO, MC, SC and FY represent the original ensemble, Grell, Low-level Omega, moisture convergence, Kain-Fritsch and Firefly closures. Categorical verification: Mean bias score versus precipitation thresholds (0.254, 2.54, 6.53, 12.7, 19.05, 25.4, 38.1, 50.8 mm) for South America for a set of 30 forecasts of 24-h accumulated precipitation for 120h in advance for (A) January 2008. Blue line represents simulations using FY weight method and red line the original EN. Blue bars indicate significance test from the bootstrap method (Hamill, 1999). FIREFLY METHOD Pseudo code EN mean Firefly begin Objective function f(x), x=(x1,..., xd)T PS wAS PAS wGR PGR wKF PKF wMC PMC wLO PLO Generate initial population of fireflies xi(i=1, 2,..., n) T Light intensity Ii at xi is determined by f(xi) J (W ) min P PS Define light absorption coefficient γ while (t < MaxGeneration)(Number of iterations) for i = 1 : n all n fireflies for j = 1 : d loop over all d dimensions if (Ij > Ii), Move firefly i towards j: end if Ir a) b) Diurnal cycle: Mean diurnal cycle of Q1 for January 2008 over northwest South America: a) ensemble mean and b) Firefly 2 0 1 xi x i ( x j xi ) rand 2 1 r 2 randomness To an environment light absorption coefficient fix I I 0e I I 0e r 2 Ir i • • • • Grell (GR) Arakawa & Schubert (AS) Kain e Fritsch (KF) Moisture convergence (MC) • Low-level omega (LO) Iterations (MaxGeneration)= 1000 Nº fireflies (n) = 20 Diurnal cycle: Hovmöller diagram of accumulated precipitation each 3h from 13 January 12:00 to 19 January 12:00 2008 mean over 10°S and 2,5°N AKNOWLEGEMENTS =O(1) => determines the convergence velocity r PM wi Pi Β 0= 1 α = 0,2 γ=1 CONCLUSION Movement of the firefly i toward firefly j (brightest) atraction 2 where 5 Numerical experiment: simulation of the diurnal cycle; 30-days of 24h forecasts January 2008 I fonte r 2 Closures used: Attractiveness varies with distance r via exp[- γr] evaluate new solutions and update light intensity end for j end for i Rank the fireflies and find the current best end while Postprocess results and visualization end Adapted of Yang (2008) I ( x) f ( x) wi M I fonte 1 r 2 The first author aknowlege the research support foundation in the state of São Paulo Fundação de Amparo à Pesquisa do Estado de São Paulo, FAPESP Results Results showed showed high high sensitivity sensitivity of of model model BRAMS BRAMS to to Firefly Firefly algorithm algorithm method, method, with with impacts impacts over over diurnal diurnal cycle cycle of of precipitation, precipitation, heatig heatig profiles profiles and and improvement improvement of of the the skill skill of of BRAMS BRAMS REFERENCES Grell, G. A. and Dévényi, D. A generalized approach to parameterizing convection combining ensemble and data assimilation techniques. Geophys. Res. Lett., v. 29, no. 14, 2002 Luz, E. F. P et al. Conceitualização do algoritmo vagalume e sua aplicação na estimativa de condição inicial de calor. In: IX Workshop do Curso de Computação Aplicada do INPE, 2009 Yang, X. Nature-Inspired Metaheuristic Algorithms, Cambridge, 2008
Documentos relacionados
Downscaling of South America present climate driven by 4
Eta Model driven by four members of an ensemble of the UK Met Office Hadley Centre HadCM3 global model. The Eta Model has been modified with the purpose of performing long-term decadal integrations...
Leia mais